DataTaunew | comments | leaders | submitlogin
Warning signs in experiment design (norvig.com)
12 points by jcbozonier 3634 days ago | 3 comments


3 points by tfturing 3634 days ago | link

Norvig proves once again that he deserves to be one of the research heads at Google!

-----

1 point by mwexler 3631 days ago | link

I too admire Norvig's writing, but I wish he would put dates on them. I never know if I'm finding something old or new, updated or out of date. I know, much of what he writes may be evergeen, but still, the world moves on and dates are helpful, if for no other reason than to properly cite the work.

-----

1 point by tfturing 3631 days ago | link

Yeah, he could have made some explicit citations to make credit more obvious. My guess is that he wanted to put his own flavor of the tremendous work by others that was included in the bibliography. In regards to the dates, they are all included in the works he did cite or the bibliography. The world moves on but that hasn't stopped people from making the same mistakes (misusing the p-value, mistaking correlation as causation). Another thing I would add to Norvig's list is the use of the word "average" in statistics. One might use that word misleadingly and just have it describe the mean when one's studies should also include the median to see if the data is skewed (a good example of this is in Darrell Huff's 1954 book). Long story short, good statistical practice will never be out of date.

-----




RSS | Announcements